argilla
argilla copied to clipboard
Add text2text example (e.g., text summarisation)
Add the text summarisation fine-tuning tutorial similar to sentiment classifier fine-tuning tutorial:
https://rubrix.readthedocs.io/en/stable/tutorials/06-labeling-finetuning.html#3.-Fine-tune-the-pre-trained-mode
I'd like to work on this, however, it seems the URL provided is outdated/broken.
Hi @iakhil ! You are right!
I think the best for this issue is to add a text summarization example here:
https://rubrix.readthedocs.io/en/stable/guides/task_examples.html#Text2Text-(Experimental)
The idea is use text summarization pretrained pipeline from Huggingface (https://huggingface.co/transformers/main_classes/pipelines.html#summarizationpipeline) and add the example in this notebook (after the machine translation one):
https://github.com/recognai/rubrix/blob/master/docs/guides/task_examples.ipynb
For example:
summarizer = pipeline("summarization")
# return three predictions
predictions = summarizer("A LONG TEXT....", num_return_sequences=3)
# log three predictions in rubrix
# follow the machine translation example
Hi @iakhil , did you manage to take a look at this? Let us know if you need support
Hey! In order to log predictions in Rubrix, the Docker-contained web app should be running, right? When I try to run it using the first method, I get the following error:
E:\rubrix>wget -O docker-compose.yml https://raw.githubusercontent.com/recognai/rubrix/master/docker-compose.yaml && docker-compose up converted 'https://raw.githubusercontent.com/recognai/rubrix/master/docker-compose.yaml' (ASCII) -> 'https://raw.githubusercontent.com/recognai/rubrix/master/docker-compose.yaml' (UTF-8) --2021-11-06 02:35:07-- https://raw.githubusercontent.com/recognai/rubrix/master/docker-compose.yaml Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.109.133, 185.199.110.133, 185.199.108.133, ... Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.109.133|:443... connected. Unable to establish SSL connection.
How should I go about it?
Hi @iakhil
There are some troubles with SSL certificates downloading the docker-compose.yml file.
You can just copy the file content and create it locally. Then launch the docker-compose up
command.
I hope this helps you
Thanks! I got it running. Are we supposed to have three predictions (as denote by the value of num_return_sequences
)? In Rubrix, I can see only one prediction for both machine translation and text summarization.
Hi @iakhil, sorry for the late reply. In principle you should be able to view the 3 predictions using the arrows on both the annotation and explore mode. Would you mind to share a screenshot with the records for which you don't see the predictions?
Earlier, I had connected to the server locally using python -m rubrix.server
but now it doesn't seem to work. I get an error that says ERROR: Application startup failed. Exiting.
Any way to fix this?
This issue is stale because it has been open for 30 days with no activity.
This issue was closed because it has been inactive for 14 days since being marked as stale.
Hey! Can I pick this issue under Hacktoberfest ? if its still pending?
Absolutely! It's still pending
@krishnajalan @iakhil https://docs.argilla.io/en/latest/guides/tasks/text_generation.html#Text-Summarization